DISTRIBUTION APPROXIMATION, ROTHS THEOREM, AND LOOKING FOR INSECTS - PowerPoint PPT Presentation

DISTRIBUTION APPROXIMATION, ROTH’S THEOREM, AND LOOKING FOR INSECTS IN SHIPPING CONTAINERS G EOFFREY D ECROUEZ U NIVERSITY OF M ELBOURNE P ETER H ALL U M ELBOURNE & UC D AVIS 1

B ACKGROUND When undertaking normal approximation to the distribution of a statistic such as the es- timator of a binomial proportion, where the sampling distribution is lattice, the lack of smoothness of the true distribution often has to be taken into account. This difficulty motivates the continuity correction, and also the fiducial approach taken by Clopper and Pearson (1934) and Sterne (1954) to estimating a binomial proportion. There is also a very large, more recent literature discussing methodology for solving prob- lems such as constructing confidence intervals for the difference or sum of two binomial proportions. Important contributions to that literature include the work of Duffy and San- ter (1987), Lee et al. (1997), Agresti and Caffo (2000), Brown et al. (2001, 2002), Zhou et al. (2001), Price and Bonnett (2004), Brown and Li (2005), Borkowf (2006), Roths and Tebbs (2006), Wang (2010) and Zieli´ nski (2010). 2

C ASES WHERE SAMPLE SIZES ARE OPEN TO CHOICE In problems where we wish to construct a confidence interval for, or test for, the differ- ence between two binomial proportions, the existing literature is founded on the premiss that the sample sizes are somehow chosen in advance. In such cases the error due to dis- continuity is generally of the same size as the error arising from issues such as skewness, kurtosis etc. However, if the sample sizes can be chosen by the experimenter, then there is considerable scope for selecting them so that the error due to discontinuity is of second order, even for one-sided procedures. This talk was motivated by a problem arising when assessing the risk that environmen- tal contaminants, in particular foreign species of insects, cross a frontier. In that setting it was desired to construct a confidence interval for the sum of two unknown binomial proportions. The experimenters were able to choose the respective sample sizes, within reasonable limits. We shall show that, in this setting, substantial reductions in the cover- age error of a confidence interval, or the level inaccuracy of a hypothesis test, are possible. 3

T HEORETICAL BACKGROUND θ denote an estimator of an unknown θ , and assume that T = n 1 / 2 (ˆ Let ˆ θ − θ ) /σ is asymp- totically normal N (0 , 1) , where n denotes sample size. Under standard assumptions, P ( T ≤ x ) = Φ( x ) + n − 1 / 2 P ( x ) φ ( x ) + o n − 1 / 2 � � , (1) uniformly in x , where Φ and φ are the standard normal distribution and density functions, respectively, and P is an even, quadratic polynomial. For example, if ˆ θ denotes the mean of a sample of size n from a population with mean θ , and if data from the population have finite third moment and a nonlattice distribution, then (1) holds with 1 − x 2 � P ( x ) = 1 � 6 β , where β is the standardised skewness of the population. On the other hand, still in the case of the mean for a population with finite third moment, if the population is lattice then an extra, discontinuous term has to be added to (1). 4

T HEORETICAL BACKGROUND – (2) The extra term reflects the discrete continuity correction that statisticians are obliged to use when approximating a lattice distribution, for example a binomial or Poisson distribution, by the smooth normal distribution: P ( T ≤ x ) = Φ( x ) + n − 1 / 2 P ( x ) φ ( x ) + n − 1 / 2 d n ( x ) φ ( x ) + o n − 1 / 2 � � . Here d n ( x ) = ( e 0 /σ ) ψ n ( x ) denotes the discontinuous term in the Edgeworth expansion, e 0 is the maximal span of the lattice, σ 2 is the population variance, ( x − ξ n ) σ n 1 / 2 � σ n 1 / 2 � � 1 � � � � � ψ n ( x ) = ψ e 0 , ξ n = e 0 2 − ψ ( n x 0 /e 0 ) , ψ ( x ) = ⌊ x ⌋ − x + 1 2 , ⌊ x ⌋ is the largest integer not strictly exceeding x , and it is assumed that all points of support in the distribution (of which θ is the mean) have the form x 0 + ν e 0 for an integer ν . 5

M ODEL FOR MULTI - SAMPLE CASES Our results and methodology are available only for multi-sample problems, which we now introduce. Let X ji , for 1 ≤ i ≤ n j and j = 1 , . . . , k , denote independent random variables. Assume that each X ji has a nondegenerate lattice distribution, depending on j but not on i and with maximal lattice edge width e j and finite third moment. Suppose too that k ≥ 2 . Put ¯ X j = n − 1 i X ji , µ j = E ( X ji ) , σ 2 � j = var( X ji ) and j k ¯ � S = X j . (2) j =1 The model (2) includes cases of apparently greater generality, for example where signed weights are incorporated in the series, since the absolute values of the weights can be incorporated into (2) by modifying the lattice edge widths, and negative signs can be ad- dressed by reflecting the summand distributions. In particular, in the case k = 2 the model includes the cases of a sum or difference of estimators of binomial proportions. 6

P ROPERTIES OF THE MODEL Since third moments are finite then, if the distributions of X 11 , . . . , X k 1 were to satisfy a er, we could express the distribution of S in a smoothness condition, such as that of Cram´ one-term Edgeworth expansion: � S − E ( S ) � = Φ( x ) + n − 1 / 2 1 1 − x 2 � n − 1 / 2 � � � P (var S ) 1 / 2 ≤ x 6 β φ ( x ) + o , (2) Here we take n = n 1 + . . . + n k to be the asymptotic parameter, and β = β ( n ) = n 1 / 2 E ( S − ES ) 3 j n − 2 n 1 / 2 � j E ( X j 1 − EX j 1 ) 3 = j n − 1 (var S ) 3 / 2 ( � j var X j 1 ) 3 / 2 is a measure of standardised skewness. Under our assumptions it is bounded as n → ∞ . However, in general (2) does not hold in the lattice-valued case. One way to see this is to consider the case k = 2 and n 1 = n 2 . Here the need to include an extra, discontinuous term, of size n − 1 / 2 , persists even in the asymptotic limit. 7

0.95 0.9 0.85 0.8 0.75 1 0.7 sqrt(2) 10 20 30 40 50 60 70 80 − 1 Figure 1:

P ROPERTIES OF THE MODEL – (2) We assume that each E | X j 1 | 3 < ∞ ; that each X j 1 is distributed on a lattice x j + ν e j , for integers ν , where e j is the maximal lattice edge width; and that the sample size ratios n j 1 /n j 2 are all bounded away from zero and infinity as n → ∞ . Recall the following formula, from the previous slide: � S − E ( S ) � = Φ( x ) + n − 1 / 2 1 1 − x 2 � n − 1 / 2 � � � P (var S ) 1 / 2 ≤ x 6 β φ ( x ) + o . (2) In the first part of the theorem below we impose the following condition on at least one of the ratios ρ j 1 j 2 = ( e j 2 n j 1 ) / ( e j 1 n j 2 ) : As n → ∞ , n 1 / 2 | sin( ℓρ j 1 j 2 π ) | → ∞ . for each integer ℓ ≥ 1 , (3) Theorem. (i) If, for some pair j 1 , j 2 with 1 ≤ j 1 < j 2 ≤ k , ρ j 1 j 2 satisfies (3) , then the one- term Edgeworth expansion at (2) holds uniformly in x . (ii) However, if ρ j 1 j 2 equals a fixed rational number (not depending on n ) for each pair j 1 , j 2 , then the expansion at (2) fails to hold because it does not include an appropriate discontinuous term of size n − 1 / 2 . 8

C ONDITION (3) Recall condition (3): n 1 / 2 | sin( ℓρ j 1 j 2 π ) | → ∞ . for each integer ℓ ≥ 1 , (3) Condition (3) holds if ρ j 1 j 2 converges to an irrational number, ρ 0 say, as n → ∞ , since if ρ 0 is irrational then | sin( ℓρ 0 π ) | > 0 for all integers ℓ . Condition (3) also holds if ρ j 1 j 2 converges sufficiently slowly to a rational number, for example if ρ j 1 j 2 = ρ 0 + ǫ j 1 j 2 , where ρ 0 is rational and ǫ j 1 j 2 = ǫ j 1 j 2 ( n ) , which can be either positive or negative, converges to zero strictly more slowly than n 1 / 2 : n 1 / 2 | ǫ j 1 j 2 | → ∞ . ǫ j 1 j 2 → 0 , On the other hand, if ρ j 1 j 2 converges sufficiently quickly to a rational number then not only does (3) fail, but the expansion (2) does not hold. Indeed, part (ii) of the theorem as- serts that the theorem fails if ρ j 1 j 2 is a fixed rational number, for each pair ( j 1 , j 2 ) and each n . Condition (3) also holds in many cases where ρ j 1 j 2 does not converge. 9

“T YPES ” OF IRRATIONAL NUMBERS The theorem shows that, if at least one of the ratios ρ j 1 j 2 converges to an irrational num- ber as n diverges, then the discontinuous term, of order n − 1 / 2 , is actually of smaller order than n − 1 / 2 . To obtain a more concise bound on the discontinuous term we shall investigate in detail cases where one or more of the ratios ρ j 1 j 2 converge to an irrational number as n diverges. This seems to require discussion of the “type” of an irrational. If x is a real number, let � x � denote the distance from x to the nearest integer. In particular, if ⌊ x ⌋ is the integer part function, � x � = min { x − ⌊ x ⌋ , 1 − ( x − ⌊ x ⌋ ) } . We say that the irrational number ρ is of type η if η equals the supremum of all ζ such that lim inf p →∞ p ζ � pρ � = 0 . Properties of convergents of irrational numbers (convergents are particularly accurate ra- tional approximations, based on continued fractions) can be used to prove that the type of any given irrational number always satisfies η ≥ 1 . 10